Many atmospheric turbulence deblurring techniques estimate an inverse filter by making assumptions that constrain the mathematical spaces in which an unknown signal and convolving function must reside. Restoration of scene content after imaging through terrestrial imaging paths is an area of active experimentation and development for both real-time feature extraction and post-process data reduction. Static scenes present opportunities for algorithms that exploit the temporal diversity of the atmospheric path since motion of scene content at the image plane over multiple frames may be attributed to a randomly varying blur kernel. This allows for the estimation of inverse filters that can be used to deblur the image. However, when objects in the scene move relative to one another across multiple image frames it complicates an already computationally demanding process. Techniques to compensate for the motion of one or more features can be used, but if the image fidelity is insufficient to detect a moving feature in the first place or the number of features (e.g. fragmentation from an impact or explosion) is very large, motion compensation techniques may break down or become impractical. In this paper we explore using multiple, synchronized optical systems with sufficient spatial separation to provide the optical path turbulence diversity required by many deblurring algorithms. This reduces or eliminates many constraints on object motion when performing reconstructions. We present deblurred imagery examples from an experimental setup that leverages spatially diverse, optical path turbulence and compare the results with the traditional approach of utilizing single path, temporal diversity when performing image reconstructions. Our results demonstrate that: (1) useful deblurring is possible with a single “set” of images simultaneously collected through diverse optical paths, (2) a combination of temporal and spatial diversity of image collection can be a useful “hybrid” approach, and (3) opportunistic weighting of concurrent frames according to image quality can enhance the deblurring results.
This paper develops a State Space model of a feed-forward control system in the frequency domain, and time domain. The results of the mathematical model are implemented and the responses of the Elevation and Azimuth servo controller in a tracking telescope called a Cine-Sextant developed for the Utah Test and Training Range.
Various approaches for transporting digital video over Ethernet and SONET networks are presented. Commercial analog and digital frame grabbers are utilized, as well as software running under Microsoft Windows 2000/XP. No other specialized hardware is required. A network configuration using independent VLANs for video channels provides efficient transport for high bandwidth data. A framework is described for implementing both uncompressed and compressed streaming with standard and non-standard video. NTSC video is handled as well as other formats that include high resolution CMOS, high bit-depth infrared, and high frame rate parallel digital. End-to-end latencies of less than 200 msec are achieved.
An approach to target tracking is presented that utilizes adaptive gray-scale correlation. The algorithm is implemented in software and executed in real-time on commercial-off-the-shelf image processing hardware. The basic correlation scheme utilizes non-adaptive commercial library calls that rely on image pyramids for speed. A framework is described for implementing a general-purpose adaptive correlation capability. The approach is robust and tolerant to scenarios involving rotation, scale changes, and contrast reversal. The algorithm also solves problems associated with walk-off.
A VME based real-time control system has been developed for use in the testing of smart munition weapons systems. The testing of advanced multimunition systems requires a platform that has not only a robust and stable servo control loop, but also a data collection platform that is capable of acquiring and tagging a wide range of sensory data. The data collection scheme must be able to handle synchronous, asynchronous, and multiframe rate sensor inputs and be capable of handling changing modes of operation in real-time. To meet these requirements, a DSP platform was utilized for the servo control loops, while programmable hardware logic was utilized to allow deterministic strobing of the time and pointing information. Discussions of the imaging requirements for this application, and limitations and uncertainties involved with optical tracking measurements are presented.
Proc. SPIE. 2468, Acquisition, Tracking, and Pointing IX
KEYWORDS: Digital signal processing, Detection and tracking algorithms, Control systems, Linear filtering, Optical tracking, Servomechanisms, Missiles, Analog electronics, Feedback control, Filtering (signal processing)
This paper investigates the problem of tracking highly dynamic targets in real-time. The advantages and disadvantages of various control and interpolation techniques are discussed. Algorithms appropriate for real- time tracking are compared with algorithms that are common in applications with a priori knowledge of the trajectory. A unique command interpolation technique combined with a state feedback control structure, implemented on a commercial DSP controller, demonstrates a significant performance improvement.
A high speed digital interface has been developed to accept real time digital pixel data from high resolution CCD cameras. The interface is currently in use with both a 640 X 486 12-bit digital infrared camera operating in non-interlaced mode at 30 frames per second, and a 756 X 484 8-bit digital visible camera operating in interlaced mode at 60 fields per second. Using programmable logic, the interface is reconfigurable to accept digital data from a variety of sensors at data rates of up to 18 megabytes per second. The buffered digital data is recorded on a hard disk array consisting of up to nine individual drives, with a present capacity exceeding 5.9 gigabytes. Continuous recording is achieved by implementing a loop function on the disk array.